How kunalj101/amexpert_ml_hackathon_2019_winners_soultions works
This repository is a collection of AmExpert coupon-redemption prediction solutions, organized as several ranked, largely independent approaches rather than one shared application. Their common workflow is to derive customer, campaign, coupon, and transaction features, train binary classifiers, and—in some approaches—combine model predictions. The rank-2 Python modules show reusable feature and model components; other capabilities are evidenced mainly by notebook and file names. Those notebooks were not inspected, so their internal wiring is intentionally not inferred.
Main entry points: modelling_adding_camp_cust_more_feats_group_k_camp_xgb_lgb.ipynb Classifier Experiments · model-lgb-xgb.ipynb Classifier Experiments · 0-LGB-0 # Feat1_LGBM_Gruped_Cats_FE.ipynb Base Classifiers · generate_features_v3.py Feature Generation
Components 30
Feature Engineering
- Feature Selection Config
Rank_1_Sourabh_Jha/Codes/remove_vars_best.json - Transaction Preparation
Rank_2_Mohsin_Khan/amlib/prepare_transaction_data.py - Train/Test Preparation
Rank_2_Mohsin_Khan/amlib/prepare_train_test_data.py - Feature Generation
Rank_2_Mohsin_Khan/amlib/generate_features_v3.py - Feature Transformers
Rank_2_Mohsin_Khan/mllib/transformers.py - Historical Artifacts
Rank_2_Mohsin_Khan/mllib/artifacts.py - Artifact Preparation
Rank_2_Mohsin_Khan/amlib/prepare_artifacts.py - Neural-Net Data Prep
Rank_2_Mohsin_Khan/amlib/prepare_nn_data.py - Coupon-Item NMF
Rank_2_Mohsin_Khan/amlib/nmf_item_coupon.py - Customer/Coupon Features
Rank_3_Sahil_Verma/feature_creation_code/feats_user_coup_item.ipynb - Variable Creation
Rank_4_Pradeep_Pathak/final_variable_creation.ipynb - Feature Engineering
Rank_5_Mudit_Tiwari/Feature Engineering 5.ipynb
Model Training
- Sourabh Pipeline
Rank_1_Sourabh_Jha/Codes/amexpert.py - Model Data Prep
Rank_2_Mohsin_Khan/mllib/prepare_data.py - Model Definition
Rank_2_Mohsin_Khan/mllib/model_definition.py - Model Runner
Rank_2_Mohsin_Khan/mllib/model_runner.py - Classifier Experiments
Rank_3_Sahil_Verma/modelling/modelling_adding_camp_cust_more_feats_group_k_camp_xgb_lgb.ipynb - Classifier Experiments
Rank_4_Pradeep_Pathak/model-lgb-xgb.ipynb - Base Classifiers
Rank_5_Mudit_Tiwari/0-LGB-0 # Feat1_LGBM_Gruped_Cats_FE.ipynb
Ensembling
- Prediction Ensembling
Rank_3_Sahil_Verma/modelling/ensembling.ipynb - Prediction Ensembling
Rank_4_Pradeep_Pathak/ensembling.ipynb - Stacking Models
Rank_5_Mudit_Tiwari/1-LOG-0 # Stacker logit.ipynb - Neural-Net Stacker
Rank_5_Mudit_Tiwari/1-NN-0 # Stacker NN.ipynb
Model Utilities
- Cross-Validation Estimator
Rank_3_Sahil_Verma/custom_estimator.py - Categorical Encoding
Rank_3_Sahil_Verma/encoding.py - Cross-Validation Estimator
Rank_4_Pradeep_Pathak/ml_modules/custom_estimator.py - Categorical Encoding
Rank_4_Pradeep_Pathak/ml_modules/encoding.py - Cross-Validation Estimator
Rank_5_Mudit_Tiwari/custom_fold_generator.py
Other components
- Competition Data
- Submission Example
Rank_3_Sahil_Verma/sample_submission_Byiv0dS.csv
Connections 27
- Competition Data → Sourabh Pipeline: provides data
- Feature Selection Config → Sourabh Pipeline: filters variables
- Competition Data → Transaction Preparation: provides data
- Transaction Preparation → Train/Test Preparation: prepares transactions
- Feature Generation → Feature Transformers: uses transforms
- Feature Generation → Historical Artifacts: uses history
- Artifact Preparation → Historical Artifacts: prepares artifacts
- Neural-Net Data Prep → Model Data Prep: prepares model data
- Coupon-Item NMF → Feature Generation: feature capability
- Model Data Prep → Model Definition: provides data
- Model Definition → Model Runner: defines model
- Competition Data → Customer/Coupon Features: provides data
- Customer/Coupon Features → Classifier Experiments: creates features
- Classifier Experiments → Cross-Validation Estimator: uses estimator
- Classifier Experiments → Categorical Encoding: uses encoding
- Classifier Experiments → Prediction Ensembling: model predictions
- Competition Data → Variable Creation: provides data
- Variable Creation → Classifier Experiments: creates variables
- Classifier Experiments → Cross-Validation Estimator: uses estimator
- Classifier Experiments → Categorical Encoding: uses encoding
- Classifier Experiments → Prediction Ensembling: model predictions
- Competition Data → Feature Engineering: provides data
- Feature Engineering → Base Classifiers: creates features
- Base Classifiers → Cross-Validation Estimator: uses fold strategy
- Base Classifiers → Stacking Models: base predictions
- Base Classifiers → Neural-Net Stacker: base predictions
- Prediction Ensembling → Submission Example: produces predictions